Lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis

Abstract

The invention discloses an auxiliary lung cancer examination tool based on biological tissue fibrous structure morphological analysis, which is characterized in that the morphological condition of a biological tissue fibrous structure in an extracellular matrix is analyzed through multi-photon microscopic imaging, and an auxiliary lung cancer examination model is established; morphological multivariate parameter quantitative characterization is performed on an elastic fiber image and a collagenous fiber image, and then the similarity degree of the two fibers on multiple morphological characteristics is comprehensively characterized on the basis, so that a normalized similarity index map is generated. According to the method, the distribution information of the elastic fibers and the collagenous fibers in the lung tissue extracellular matrix can be rapidly extracted, the lung tumor evolution level is judged according to the difference of morphological distribution similarity conditions of the two fibrous structures of a normal person and a lung cancer patient, and the method serves as a novel examination method provided by the invention. The method has great clinical transformation potential, can more accurately analyze the morphological condition of the fibrous structure of the biological tissue, and has greater application potential compared with the traditional method.

Description

Technical Field

[0001] The present invention belongs to the technical field of image analysis and disease detection of biological tissues, and particularly relates to a lung cancer auxiliary examination tool based on morphological analysis of fibrous structures of biological tissues. Background Art

[0002] Lung cancer is the most frequently occurring cancer globally in recent years and continues to rank first among the causes of cancer death. There are mainly two histological subtypes of lung cancer: small cell lung cancer and non-small cell lung cancer (NSCLC), with the latter accounting for approximately 85% of total cases, being less invasive and having a slower growth rate. However, the prognosis of NSCLC patients remains poor, with more than half of the patients dying within one year after diagnosis and a 5-year survival rate of less than 20%. Therefore, accurate staging is crucial for clinical treatment and scientific research. Although the examination of stained sections of biopsy specimens has always been the gold standard for NSCLC staging, operations such as biopsies and surgical resections involving the confirmation of tumor boundaries still have disadvantages such as complex sampling, long time consumption, and heavy dependence on doctors' experience.

[0003] Currently, it is known that various carcinogenesis processes are often accompanied by tissue sclerosis, including lung cancer. Microscopically, the remodeling of the extracellular matrix during cancer evolution has gradually become an important marker for cancer diagnosis, staging, and grading; among them, the morphological changes of some fibrous structures of the extracellular matrix can also reflect the changes in macroscopic mechanical properties, thereby indirectly reflecting the evolution of lung tumors. According to this characteristic, multiphoton microscopy is used to perform morphological analysis on the fibrous structures (mainly elastic fibers and collagen fibers) in the extracellular matrix. According to the similarity degree of the two types of fibers, the lesion level of lung cancer can be indirectly inferred.

[0004] Multiphoton microscopy (MPM for short) has the advantages of deep imaging penetration depth and high optical three-dimensional resolution, and has been widely used. MPM technology can perform two-photon excited fluorescence (TPEF) microscopy imaging and second harmonic generation (SHG) imaging on the elastic fibers and collagen fibers of the extracellular matrix respectively, with the advantages of label-free and fast imaging. Analyzing the imaging results can obtain information related to the evolution of lung cancer.

[0005] The prior art has the following technical problems:

[0006] 1) The previous traditional histological diagnostic methods relied on hematoxylin and eosin stained sections of biopsy samples obtained by direct sampling. In addition to requiring invasive sampling from a living body, the time for staining and preparing the slides after sampling was relatively long, about one week; and there were often situations where the tumor spread after the diagnosis of the slides and re-sampling was required. Therefore, there is a need for an auxiliary diagnostic method that can more quickly and accurately distinguish the boundary of lung tumors; finally, traditional sampling requires thoracotomy to remove a piece of lung tissue for examination, with a relatively large window.

[0007] 2) Currently, there are emerging in vivo imaging technologies that can achieve low-invasive, rapid and accurate auxiliary diagnosis for early-stage in situ lung cancer. However, they usually involve the use of radioactive contrast agents and complex operations, and most require invasive examinations except for some endoscopic techniques, and still leave wounds in the body.

[0008] 3) Other emerging technologies also include techniques such as endobronchial ultrasound-guided fine needle aspiration (EBUS-FNA) and endoscopic ultrasound-guided fine needle aspiration through the esophagus (EUS-FNA). Besides their relatively limited sensitivity, they require additional time for training.

[0009] 4) Further, after biopsy, it may be necessary to resect early-stage in situ tumors. Although video-assisted thoracoscopic surgery (VATS) and robotic-assisted thoracic surgery (RATS) can achieve minimally invasive surgery, the potential risks during the operation are relatively large, and on the premise of abandoning traditional equipment, the newly introduced robotic consumables have high costs.

[0010] 5) Currently, the uses of endoscopes are still relatively single, often highly specialized devices applicable to single diseases or organs, with high costs, and lack general-purpose auxiliary diagnostic methods and devices applicable to similar diagnostic and treatment principles such as tumor sclerosis to reduce costs. Summary of the Invention

[0011] In view of the deficiencies of the prior art, the present invention provides a lung cancer auxiliary examination tool based on the morphological analysis of the fibrous structure of biological tissues. This method uses an MPM system to image the extracellular matrix in lung tissues, performs multi-parameter characterization of the morphological features of elastic fibers and collagen fibers in the multi-photon microscopy imaging results, quantifies the similarity degree of the two types of fibers based on the characterization results at the pixel level, and further establishes a classification model according to the statistical indicators of the normalized similarity index map to achieve the auxiliary examination of lung cancer. This method avoids the disadvantages of the traditional method for diagnosing lung cancer, such as lack of quantification, time-consuming and laborious, and dependence on experience, and provides a new method for the in vivo diagnosis and related research of human lung tumors.

[0012] The technical solution adopted by the present invention is as follows:

[0013] The present invention discloses a lung cancer auxiliary examination tool based on the morphological analysis of the fibrous structure of biological tissues, characterized in that it analyzes the morphological situation of the fibrous structure of biological tissues in the extracellular matrix through multi-photon microscopy imaging and establishes a lung cancer auxiliary examination model. The examination tool includes the following steps:

[0014] 1) By multi - photon microscopy, two - photon excited fluorescence imaging and second - harmonic generation imaging are respectively performed on elastic fibers and collagen fibers in the extracellular matrix of human lung tissue to obtain multi - photon microscopy images, including elastic fiber images and collagen fiber images;

[0015] 2) Morphological multi - parameter quantitative characterization is respectively performed on the elastic fiber images and collagen fiber images obtained in step 1). The morphological multi - parameters include local density, spatial orientation, direction variance, curvature, and diameter, so as to extract the morphological features of elastic fibers and collagen fibers at the pixel - level accuracy. For the elastic fiber images, a series of original morphological parameter maps of elastic fibers are generated, including: elastic fiber local density parameter map, elastic fiber spatial orientation parameter map, elastic fiber direction variance parameter map, elastic fiber curvature parameter map, and elastic fiber diameter parameter map. For the collagen fiber images, a series of original morphological parameter maps of collagen fibers are generated, including: collagen fiber local density parameter map, collagen fiber spatial orientation parameter map, collagen fiber direction variance parameter map, and collagen fiber curvature parameter map;

[0016] 3) Image segmentation is respectively performed on the elastic fiber images and collagen fiber images obtained in step 1). The image segmentation results are respectively denoted as m1 and m2. In the segmentation results, the pixels representing elastic fibers or collagen fibers are 1, and the pixels representing the background are 0;

[0017] 4) For m1, cross - channel nearest - neighbor fiber pixel retrieval is performed, that is, the nearest - neighbor non - zero pixels in m2 to all non - zero pixels in it are searched, and the distances between the corresponding pixels are calculated to generate an elastic fiber nearest - neighbor distance parameter map. At the same time, the values of the morphological parameters at the nearest - neighbor non - zero pixels in m2 are assigned to the non - zero pixels in m1, so as to generate an elastic fiber nearest - neighbor pixel morphological parameter map involving the parameters in step 2), including: elastic fiber nearest - neighbor pixel local density parameter map, elastic fiber nearest - neighbor pixel spatial orientation parameter map, elastic fiber nearest - neighbor pixel direction variance parameter map, and elastic fiber nearest - neighbor pixel curvature parameter map;

[0018] 5) For m2, cross - channel nearest - neighbor fiber pixel retrieval is performed, that is, the nearest - neighbor non - zero pixels in m1 to all non - zero pixels in it are searched, and the distances between the corresponding pixels are calculated to generate a collagen fiber nearest - neighbor distance parameter map. At the same time, the values of the morphological parameters at the nearest - neighbor non - zero pixels in m1 are assigned to the non - zero pixels in m2, so as to generate a collagen fiber nearest - neighbor pixel morphological parameter map involving the parameters in step 2), including: collagen fiber nearest - neighbor pixel local density parameter map, collagen fiber nearest - neighbor pixel spatial orientation parameter map, collagen fiber nearest - neighbor pixel direction variance parameter map, and collagen fiber nearest - neighbor pixel curvature parameter map;

[0019] 6) According to the morphological parameter map of the nearest neighbor pixels of elastic fibers obtained in step 4), perform a differential process on it and the corresponding original morphological parameter map of elastic fibers obtained in step 2) to obtain the morphological parameter difference map of the nearest neighbor pixels of elastic fibers;

[0020] 7) According to the morphological parameter map of the nearest neighbor pixels of collagen fibers obtained in step 5), perform a differential process on it and the corresponding original morphological parameter map of collagen fibers obtained in step 2) to obtain the morphological parameter difference map of the nearest neighbor pixels of collagen fibers;

[0021] 8) Perform pairwise fusion weighted by the image segmentation result on the nearest neighbor distance parameter map of elastic fibers obtained in step 4) and the nearest neighbor distance parameter map of collagen fibers obtained in step 5) to obtain the difference fusion map of the nearest neighbor pixel distance parameter. And for the morphological parameter difference map of the nearest neighbor pixels of elastic fibers obtained in step 6) and the morphological parameter difference map of the nearest neighbor pixels of collagen fibers obtained in step 7), perform pairwise fusion weighted by the image segmentation result according to the corresponding morphological parameters to obtain a series of difference fusion maps of the nearest neighbor pixel morphological parameters. The difference fusion map of the nearest neighbor pixel distance parameter and the series of difference fusion maps of the nearest neighbor pixel morphological parameters are collectively referred to as the difference fusion map of the nearest neighbor pixel parameters;

[0022] 9) Perform normalization scoring on the difference fusion map of the nearest neighbor pixel parameters obtained in step 8) to obtain a series of normalized difference fusion score maps of the nearest neighbor pixel parameters, including: the normalized difference fusion score map of the nearest neighbor distance parameter, the normalized difference fusion score map of the local density parameter of the nearest neighbor pixel, the normalized difference fusion score map of the spatial orientation parameter of the nearest neighbor pixel, the normalized difference fusion score map of the direction variance parameter of the nearest neighbor pixel, and the normalized difference fusion score map of the curvature parameter of the nearest neighbor pixel. In the map, the pixels with scores closer to 0 indicate that the corresponding morphological features of elastic fibers and collagen fibers are more different at that position, and the pixels with scores closer to 1 indicate that the corresponding morphological features of elastic fibers and collagen fibers are more similar at that position. Then multiply the corresponding pixels of all the normalized difference fusion score maps of the nearest neighbor pixel parameters, and finally obtain the normalized similarity index map reflecting the similarity degree of the morphology of elastic fibers and collagen fibers;

[0023] 10) Calculate the full-stack average value, full-stack standard deviation value, and full-stack depth direction variation value of the non-background pixel part of the normalized similarity index map respectively, and indirectly reflect the hardness change of the lung tissue through the combination of the three values, so as to realize the extraction of information related to the evolution of lung cancer.

[0024] As a further improvement, in step 6) of the present invention, according to the morphological parameter map of the nearest neighbor pixels of elastic fibers obtained in step 4), a differential treatment is performed on it and the corresponding original morphological parameter map of elastic fibers obtained in step 2). Specifically: a differential treatment is performed between the local density parameter map of the nearest neighbor pixels of elastic fibers and the local density parameter map of elastic fibers to obtain a local density parameter difference map of the nearest neighbor pixels of elastic fibers; a differential treatment is performed between the spatial orientation parameter map of the nearest neighbor pixels of elastic fibers and the spatial orientation parameter map of elastic fibers to obtain a spatial orientation parameter difference map of the nearest neighbor pixels of elastic fibers; a differential treatment is performed between the direction variance parameter map of the nearest neighbor pixels of elastic fibers and the direction variance parameter map of elastic fibers to obtain a direction variance parameter difference map of the nearest neighbor pixels of elastic fibers; a differential treatment is performed between the curvature parameter map of the nearest neighbor pixels of elastic fibers and the curvature parameter map of elastic fibers to obtain a curvature parameter difference map of the nearest neighbor pixels of elastic fibers. The local density parameter difference map of the nearest neighbor pixels of elastic fibers, the spatial orientation parameter difference map of the nearest neighbor pixels of elastic fibers, the direction variance parameter difference map of the nearest neighbor pixels of elastic fibers, and the curvature parameter difference map of the nearest neighbor pixels of elastic fibers obtained by the differential treatment are collectively referred to as the morphological parameter difference map of the nearest neighbor pixels of elastic fibers. Among them, for the morphological parameter of spatial orientation, when the difference result is between 90° and 180°, its supplementary angle is taken as the differential result.

[0025] As a further improvement, in step 7) of the present invention, according to the morphological parameter map of the nearest neighbor pixels of collagen fibers obtained in step 5), a differential treatment is performed on it and the corresponding original morphological parameter map of collagen fibers obtained in step 2). Specifically: a differential treatment is performed between the local density parameter map of the nearest neighbor pixels of collagen fibers and the local density parameter map of collagen fibers to obtain a local density parameter difference map of the nearest neighbor pixels of collagen fibers; a differential treatment is performed between the spatial orientation parameter map of the nearest neighbor pixels of collagen fibers and the spatial orientation parameter map of collagen fibers to obtain a spatial orientation parameter difference map of the nearest neighbor pixels of collagen fibers; a differential treatment is performed between the direction variance parameter map of the nearest neighbor pixels of collagen fibers and the direction variance parameter map of collagen fibers to obtain a direction variance parameter difference map of the nearest neighbor pixels of collagen fibers; a differential treatment is performed between the curvature parameter map of the nearest neighbor pixels of collagen fibers and the curvature parameter map of collagen fibers to obtain a curvature parameter difference map of the nearest neighbor pixels of collagen fibers. The local density parameter difference map of the nearest neighbor pixels of collagen fibers, the spatial orientation parameter difference map of the nearest neighbor pixels of collagen fibers, the direction variance parameter difference map of the nearest neighbor pixels of collagen fibers, and the curvature parameter difference map of the nearest neighbor pixels of collagen fibers obtained by the differential treatment are collectively referred to as the morphological parameter difference map of the nearest neighbor pixels of collagen fibers. Among them, for the morphological parameter of spatial orientation, when the difference result is between 90° and 180°, its supplementary angle is taken as the differential result.

[0026] As a further improvement, in step 8) of the present invention, according to the morphological parameter difference map of the nearest neighbor pixels of elastic fibers obtained in step 6) and the morphological parameter difference map of the nearest neighbor pixels of collagen fibers obtained in step 7), pairwise fusion weighted by the image segmentation result is performed according to the corresponding morphological parameters. Specifically: the local density parameter difference map of the nearest neighbor pixels of elastic fibers is fused with the local density parameter difference map of the nearest neighbor pixels of collagen fibers to obtain the local density parameter difference fusion map of the nearest neighbor pixels; the spatial orientation parameter difference map of the nearest neighbor pixels of elastic fibers is fused with the spatial orientation parameter difference map of the nearest neighbor pixels of collagen fibers to obtain the spatial orientation parameter difference fusion map of the nearest neighbor pixels; the direction variance parameter difference map of the nearest neighbor pixels of elastic fibers is fused with the direction variance parameter difference map of the nearest neighbor pixels of collagen fibers to obtain the direction variance parameter difference fusion map of the nearest neighbor pixels; the curvature parameter difference map of the nearest neighbor pixels of elastic fibers is fused with the curvature parameter difference map of the nearest neighbor pixels of collagen fibers to obtain the curvature parameter difference fusion map of the nearest neighbor pixels.

[0027] As a further improvement, when performing pairwise fusion weighted by the image segmentation result according to the corresponding morphological parameters on the morphological parameter difference map of the nearest neighbor pixels of elastic fibers obtained in step 6) and the morphological parameter difference map of the nearest neighbor pixels of collagen fibers obtained in step 7) in step 8) of the present invention, the specific way of weighting by the image segmentation result is as follows:

[0028]

[0029] where M FP is the morphological parameter difference fusion map of the nearest neighbor pixels of the corresponding morphological parameters of elastic fibers and collagen fibers, M SP1 is the morphological parameter difference map of the nearest neighbor pixels of any morphological parameter of elastic fibers, M SP2 is the morphological parameter difference map of the nearest neighbor pixels of the corresponding morphological parameter of collagen fibers, and m1 and m2 are the image segmentation results of the elastic fiber image and the collagen fiber image respectively.

[0030] As a further improvement, in step 9) of the present invention, for the nearest neighbor pixel parameter difference fusion map obtained in step 8), normalization scoring is performed. For the nearest neighbor distance parameter difference fusion map M FD , the following expression is used:

[0031]

[0032] where S D is the normalized nearest neighbor distance parameter difference fusion score map, D uplimitis a reference value related to the fused atlas of nearest neighbor pixel distance parameters or the elastic fiber diameter, taken as the maximum diameter value in the elastic fiber diameter parameter atlas, where e is the base of the natural logarithm; for the fused atlas M of the nearest neighbor pixel local density parameter differences FL , the following expression is used:

[0033]

[0034] where S L is the normalized fused score atlas of nearest neighbor pixel local density parameter differences; for the fused atlas M of the nearest neighbor pixel spatial orientation parameter differences FO , the following expression is used:

[0035] S O = cos(M FO );

[0036] where S O is the normalized fused score atlas of nearest neighbor pixel spatial orientation parameter differences; for the fused atlas M of the nearest neighbor distance direction variance parameter differences or the nearest neighbor distance curvature parameter differences FCD , the following expression is used:

[0037]

[0038] where S CD is the normalized fused score atlas of nearest neighbor pixel direction variance parameter differences or the normalized fused score atlas of nearest neighbor pixel curvature parameter differences, and C is an empirical coefficient, generally taken as 4.

[0039] As a further improvement, in step 9) of the present invention, the calculation method of the full-stack depth direction variation value is: first calculate the average value of all non-background pixels in each two-dimensional atlas layer of the normalized similarity index atlas, and then calculate the standard deviation of these average values.

[0040] As a further improvement, in step 9) of the present invention, for the extraction of information related to lung cancer evolution, the support vector machine method is further used as an auxiliary inspection tool for the lung tumor boundary.

[0041] The present invention uses multi-photon microscopy to achieve high-resolution and fast in-vivo inspection. By morphologically quantitatively characterizing the images of elastic fibers and collagen fibers in the extracellular matrix, the similarity degree of the two fibers can indirectly reveal the evolution information of lung tumors by reflecting the change in matrix hardness. Using the established lung cancer auxiliary inspection tool to assist doctors in the inspection can further strengthen the pathological understanding of human lung cancer.

[0042] Due to the application of the above technical solutions, the present invention has the following advantages compared with the prior art solutions:

[0043] The present invention discloses a lung cancer auxiliary examination tool for morphological analysis of fibrous structures in biological tissues based on MPM, which is used to analyze the morphological conditions of fibrous structures in biological tissues in the outer plexiform layer of the extracellular matrix and diagnose abnormalities in macroscopic hardness and microscopic fibrous structure remodeling caused by lung cancer. The present invention can quantitatively characterize multiple morphological parameters of elastic fiber and collagen fiber images morphologically, and then comprehensively characterize the similarity degree of the two types of fibers in multiple morphological features on this basis, so as to generate a normalized similarity index map. An auxiliary examination model for lung cancer is established according to the statistical features related to the normalized similarity index map, and this model can automatically perform lung cancer examinations. The present invention can quickly extract the distribution information of elastic fibers and collagen fibers in the extracellular matrix of lung tissues, and judge the evolution level of lung tumors according to the difference in the similarity of the morphological distributions of the two types of fibrous structures between normal people and lung cancer patients, which is a new examination method proposed by the present invention. This method has great potential for clinical transformation, can more accurately analyze the morphological conditions of fibrous structures in biological tissues, and has greater application potential than traditional methods.

[0044] 1. Different from the traditional histological diagnosis method that relies on hematoxylin and eosin staining sections of biopsy samples by direct sampling in the past, the present invention selects multi-photon microscopy imaging of human lung tissues to perform rapid, label-free, and high-resolution imaging of the extracellular matrix, and diagnoses lung cancer by detecting the microscopic remodeling of the extracellular matrix, so as to realize the discriminant analysis of the lung cancer evolution process. Due to the characteristics of MPM imaging, it can achieve non-destructive three-dimensional imaging with a certain penetration depth of in-situ organs. In comparison, it can obtain more detailed information on tissue changes during the lung cancer evolution process.

[0045] 2. Based on the MPM imaging results of two fibrous structures in the extracellular matrix, the present invention extracts and analyzes the morphological characteristics of elastic fibers and collagen fiber structures, and obtains the tissue distribution of the two fibrous structures in morphological characteristics based on this. The similarity degree of the two types of fibers can indirectly reflect the degree of tumor sclerosis, thereby further revealing the development status of lung cancer. Using MPM images for morphological analysis breaks the complex operations of previous techniques such as medical images using radioactive contrast agents, and can empower traditional techniques such as traditional thoracoscopic biopsies, realizing low-damage, rapid, accurate, and simple auxiliary diagnosis of early in-situ lung cancer.

[0046] 3. The present invention innovatively proposes a method for quantitatively characterizing the similarity index, which indirectly reflects the evolution of lung cancer by reflecting the change in the hardness of lung tissue. This parameter synthesizes a variety of morphological features, including the nearest distance of fibers related to the pixel intensity and distribution of fibrous structures, the local density of fibers, the spatial orientation of fibrous structures, and the direction variance (order degree) and curvature related to the tissue and distribution characteristics of fibrous structures. By synthesizing the above-mentioned various morphological features, the obtained normalized parameter can intuitively reflect the overall similarity of two fibrous structures, has strong interpretability, and can provide histological information for doctors more simply and directly without complex training and a large amount of experience accumulation.

[0047] 4. The present invention is based on three-dimensional label-free in vivo imaging analysis. Compared with the analysis based on single images and video guidance, the utilization rate of the spatial information in the depth direction of MPM images has been greatly improved, and the extraction of the extracellular matrix microscopically also provides updated information for diagnosis and treatment. This method provides a new idea for boundary recognition in the clinical in vivo sampling process and in situ tumor resection surgery, and can effectively improve the accuracy of the inspection tool for exploring the boundary of lung tumors. At the same time, this method can be technically transplanted based on traditional thoracoscopic instruments, and has good clinical transformation significance.

[0048] 5. The present invention realizes the morphological distribution similarity analysis of two fibrous structures in MPM images and applies it to the microscopic detection of lung cancer evolution. Further, the present invention can not only perform similarity analysis on two or more fibrous structures microscopically. Since the present invention reflects the macroscopic hardness based on the microscopic remodeling of the fibrous structure of the extracellular matrix, it can be extended to the exploration and application of more cancers and diseases that conform to this law, which is of great significance for revealing the information of the change in the fibrous structure in the microscopic biological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Flow chart for constructing an auxiliary inspection tool for lung cancer based on the morphological similarity analysis of the fibrous structure of extracellular matrix biological tissue;

[0050] Figure 2 Comparison chart of the imaging results of two fibrous structures and the quantitative characterization results of morphological multi-parameters in the extracellular matrix of ex vivo healthy human lungs and NSCLC lung cancer under MPM;

[0051] Figure 3 Flow chart for quantitatively characterizing the morphological similarity index using simulated fibers;

[0052] Figure 4 Schematic diagram of the application of the classification model for similarity index analysis based on the in vivo MPM imaging results of mice. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the following provides a detailed description of the technical solutions of the present invention in conjunction with the specification drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0054] The present invention is an auxiliary lung cancer examination tool based on the morphological analysis of the fibrous structure of extracellular matrix biological tissues, and the specific steps are as Figure 1 shown:

[0055] 1) Multiphoton microscopy imaging of extracellular matrix: Use a clinical MPM instrument to image the extracellular matrix of suspected cancerous tissue parts of the human lung, and perform two-photon excited fluorescence imaging and second harmonic generation imaging on elastic fibers and collagen fibers in the extracellular matrix of human lung tissue respectively. Each of the two imaging methods occupies one channel, thereby obtaining multiphoton microscopy images, including elastic fiber images and collagen fiber images. The three-dimensional fusion reconstruction of the elastic fiber images and collagen fiber images of healthy human lungs and human lungs with NSCLC is as Figure 2 (a) shown.

[0056] 2) Morphological multi-parameter quantitative characterization is performed on the elastic fiber images and collagen fiber images obtained in step 1) respectively. The morphological multi-parameters include local density, spatial orientation, direction variance, curvature, and diameter, so as to extract the morphological characteristics of elastic fibers and collagen fibers at the pixel level accuracy. For the elastic fiber images, a series of original morphological parameter maps of elastic fibers are generated, including: elastic fiber local density parameter map, elastic fiber spatial orientation parameter map, elastic fiber direction variance parameter map, elastic fiber curvature parameter map, and elastic fiber diameter parameter map. For the collagen fiber images, a series of original morphological parameter maps of collagen fibers are generated, including: collagen fiber local density parameter map, collagen fiber spatial orientation parameter map, collagen fiber direction variance parameter map, and collagen fiber curvature parameter map, as Figure 2 (b) shown. For a more intuitive presentation, the fusion map of the morphological multi-parameter quantitative characterization results of the two-channel two-dimensional simulated fibers of the elastic fiber images and collagen fiber images is shown in Figure 3 (a), where the color-coded legend is of the hsv type for spatial orientation and of the jet type for the others.

[0057] 3) Image segmentation is performed on the elastic fiber images and collagen fiber images obtained in step 1) respectively, and the obtained image segmentation results are denoted as m1 and m2 respectively. In the segmentation results, the pixels representing elastic fibers or collagen fibers are 1, and the pixels representing the background are 0, as Figure 3 (b) the leftmost column shown.

[0058] 4) For m1, perform cross-channel nearest neighbor fiber pixel retrieval, that is, find the nearest neighbor non-zero pixels in m2 for all non-zero pixels in it, calculate the distances between the corresponding pixels, generate the elastic fiber nearest neighbor distance parameter map, and at the same time assign the values of the morphological parameters at the nearest neighbor non-zero pixels in m2 to the non-zero pixels in m1, so as to generate the elastic fiber nearest neighbor pixel morphological parameter map involving the parameters in step 2), including: elastic fiber nearest neighbor pixel local density parameter map, elastic fiber nearest neighbor pixel spatial orientation parameter map, elastic fiber nearest neighbor pixel direction variance parameter map, and elastic fiber nearest neighbor pixel curvature parameter map, thus obtaining the result shown to the right of the arrow in the upper half of Figure 3 (b).

[0059] 5) For m2, perform cross-channel nearest neighbor fiber pixel retrieval, that is, find the nearest neighbor non-zero pixels in m1 for all non-zero pixels in it, calculate the distances between the corresponding pixels, generate the collagen fiber nearest neighbor distance parameter map, and at the same time assign the values of the morphological parameters at the nearest neighbor non-zero pixels in m1 to the non-zero pixels in m2, so as to generate the collagen fiber nearest neighbor pixel morphological parameter map involving the parameters in step 2), including: collagen fiber nearest neighbor pixel local density parameter map, collagen fiber nearest neighbor pixel spatial orientation parameter map, collagen fiber nearest neighbor pixel direction variance parameter map, and collagen fiber nearest neighbor pixel curvature parameter map, thus obtaining the result shown to the right of the arrow in the lower half of Figure 3 (b).

[0060] 6) According to the morphological parameter map of the nearest neighbor pixels of the elastic fibers obtained in step 4), perform a differential processing on it with the corresponding original morphological parameter map of the elastic fibers obtained in step 2) to obtain a morphological parameter difference map of the nearest neighbor pixels of the elastic fibers. Specifically, perform an operation of subtraction first and then absolute value. That is, perform differential processing on the local density parameter map of the nearest neighbor pixels of the elastic fibers and the local density parameter map of the elastic fibers respectively to obtain a local density parameter difference map of the nearest neighbor pixels of the elastic fibers, perform differential processing on the spatial orientation parameter map of the nearest neighbor pixels of the elastic fibers and the spatial orientation parameter map of the elastic fibers to obtain a spatial orientation parameter difference map of the nearest neighbor pixels of the elastic fibers, perform differential processing on the direction variance parameter map of the nearest neighbor pixels of the elastic fibers and the direction variance parameter map of the elastic fibers to obtain a direction variance parameter difference map of the nearest neighbor pixels of the elastic fibers, perform differential processing on the curvature parameter map of the nearest neighbor pixels of the elastic fibers and the curvature parameter map of the elastic fibers to obtain a curvature parameter difference map of the nearest neighbor pixels of the elastic fibers. The local density parameter difference map of the nearest neighbor pixels of the elastic fibers, the spatial orientation parameter difference map of the nearest neighbor pixels of the elastic fibers, the direction variance parameter difference map of the nearest neighbor pixels of the elastic fibers, and the curvature parameter difference map of the nearest neighbor pixels of the elastic fibers obtained by differential processing are collectively referred to as the morphological parameter difference map of the nearest neighbor pixels of the elastic fibers. In particular, for the parameter of spatial orientation, when the result of subtraction is between 90° and 180°, directly use its supplementary angle as the differential result.

[0061] 7) According to the morphological parameter map of the nearest neighbor pixels of collagen fibers obtained in step 5), perform a differential treatment on it and the corresponding original morphological parameter map of collagen fibers obtained in step 2) to obtain the morphological parameter difference map of the nearest neighbor pixels of collagen fibers. Specifically, perform an operation of taking the difference first and then taking the absolute value, that is, perform a differential treatment on the local density parameter map of the nearest neighbor pixels of collagen fibers and the local density parameter map of collagen fibers respectively to obtain the local density parameter difference map of the nearest neighbor pixels of collagen fibers, perform a differential treatment on the spatial orientation parameter map of the nearest neighbor pixels of collagen fibers and the spatial orientation parameter map of collagen fibers to obtain the spatial orientation parameter difference map of the nearest neighbor pixels of collagen fibers, perform a differential treatment on the direction variance parameter map of the nearest neighbor pixels of collagen fibers and the direction variance parameter map of collagen fibers to obtain the direction variance parameter difference map of the nearest neighbor pixels of collagen fibers, perform a differential treatment on the curvature parameter map of the nearest neighbor pixels of collagen fibers and the curvature parameter map of collagen fibers to obtain the curvature parameter difference map of the nearest neighbor pixels of collagen fibers. The local density parameter difference map of the nearest neighbor pixels of collagen fibers, the spatial orientation parameter difference map of the nearest neighbor pixels of collagen fibers, the direction variance parameter difference map of the nearest neighbor pixels of collagen fibers, and the curvature parameter difference map of the nearest neighbor pixels of collagen fibers obtained by the differential treatment are collectively referred to as the morphological parameter difference map of the nearest neighbor pixels of collagen fibers. In particular, for the parameter of spatial orientation, when the difference result is between 90° and 180°, directly use its supplementary angle as the differential result.

[0062] 8) Perform pairwise fusion weighted by the image segmentation result on the nearest neighbor distance parameter map of elastic fibers obtained in step 4) and the nearest neighbor distance parameter map of collagen fibers obtained in step 5) to obtain the differential fusion map of the nearest neighbor pixel distance parameter. And for the morphological parameter difference map of the nearest neighbor pixels of elastic fibers obtained in step 6) and the morphological parameter difference map of the nearest neighbor pixels of collagen fibers obtained in step 7), perform pairwise fusion weighted by the image segmentation result according to the corresponding morphological parameters to obtain a series of differential fusion maps of the nearest neighbor pixel morphological parameters. The differential fusion map of the nearest neighbor pixel distance parameter and a series of differential fusion maps of the nearest neighbor pixel morphological parameters are collectively referred to as the differential fusion map of the nearest neighbor pixel parameters. When performing pairwise fusion weighted by the image segmentation result according to the corresponding morphological parameters, the specific way of weighting by the image segmentation result is as follows:

[0063]

[0064] Among them, M FP is the differential fusion map of the nearest neighbor pixel parameters of the corresponding morphological parameters of elastic fibers and collagen fibers, M SP1 is the morphological parameter difference map of the nearest neighbor pixels of any morphological parameter of elastic fibers, M SP2It is the morphological parameter difference map of the nearest neighbor pixels corresponding to the morphological parameters of collagen fibers. m1 and m2 are the image segmentation results of the elastic fiber image and the collagen fiber image respectively. The nearest neighbor pixel parameter difference fusion map is as shown in Figure 3 (c), where the color-coded legend is of the jet type.

[0065] 9) Normalize and score the nearest neighbor pixel parameter difference fusion map obtained in step 8) to obtain a series of normalized nearest neighbor pixel parameter difference fusion score maps. Specifically, different scoring strategies are used according to different morphological features, such as Figure 3 (d). A series of normalized nearest neighbor pixel parameter difference fusion score maps include: normalized nearest neighbor distance parameter difference fusion score map, normalized nearest neighbor pixel local density parameter difference fusion score map, normalized nearest neighbor pixel spatial orientation parameter difference fusion score map, normalized nearest neighbor pixel direction variance parameter difference fusion score map, and normalized nearest neighbor pixel curvature parameter difference fusion score map. Among them, the pixels with scores closer to 0 indicate that the corresponding morphological features of elastic fibers and collagen fibers are more different at that position, and the pixels with scores closer to 1 indicate that the corresponding morphological features of elastic fibers and collagen fibers are more similar at that position. Then, multiply the corresponding pixels of all normalized nearest neighbor pixel parameter difference fusion score maps to finally obtain a normalized similarity index map reflecting the morphological similarity between elastic fibers and collagen fibers, as shown in Figure 3 (e), where the color-coded legend is of the jet type. When normalizing and scoring, for the nearest neighbor distance parameter difference fusion map M FD , the following expression is used:

[0066]

[0067] where S D is the normalized nearest neighbor distance parameter difference fusion score map, D uplimit is a reference value related to the nearest neighbor pixel distance parameter fusion map or the elastic fiber diameter, taken as the maximum diameter value in the elastic fiber diameter parameter map, and e is the base of the natural logarithm. For the nearest neighbor pixel local density parameter difference fusion map M FL , the following expression is used:

[0068]

[0069] where S L is the normalized nearest neighbor pixel local density parameter difference fusion score map. For the nearest neighbor pixel spatial orientation parameter difference fusion map M FO , the following expression is used:

[0070] S O= cos(M FO )

[0071] where S O is the normalized nearest neighbor pixel spatial orientation parameter difference fusion score map; for the nearest neighbor distance direction variance parameter difference fusion map or the nearest neighbor distance curvature parameter difference fusion map, the following expression is used:

[0072]

[0073] where S CD is the normalized nearest neighbor pixel direction variance parameter difference fusion score map or the normalized nearest neighbor pixel curvature parameter difference fusion score map, C is an empirical coefficient, generally taken as 4.

[0074] 10) Calculate the full-stack average value, full-stack standard deviation value, and full-stack depth direction variation value of the non-background pixel part in the normalized similarity index map respectively. Among them, the calculation method of the full-stack depth direction variation value is: first calculate the average value of all non-background pixels in each two-dimensional map of each layer of the normalized similarity index map, and then calculate the standard deviation of these average values. The combination of the three values indirectly reflects the hardness change of the lung tissue, so as to realize the extraction of information related to the evolution of lung cancer. Further use the support vector machine method as an auxiliary inspection tool for the boundary of lung tumors. As Figure 4 (a) shows, the hardness of lung cancer tissues is significantly higher than that of healthy tissues, and Figure 4 (b) the linear regression result shows that the similarity index parameter can reflect the hardness change of the lung tissue during the development of cancer. Based on this, calculate the full-stack average value, full-stack standard deviation value, and full-stack depth direction variation value of the non-background pixel part in the normalized similarity index map respectively. Among them, the calculation method of the full-stack depth direction variation value is: first calculate the average value of all non-background pixels in each two-dimensional map of each layer of the normalized similarity index map, and then calculate the standard deviation of these average values. As Figure 4 (c) shows, the above three statistical values of the similarity index have strong discrimination among the three groups. Therefore, the combination of the three values can indirectly reflect the hardness change of the lung tissue, so as to realize the extraction of information related to the evolution of lung cancer. As Figure 4 (d) shows, the scatter plot of the three values can basically distinguish the healthy, boundary, and cancerous tissues of the in-vivo lung imaging results of mice.

[0075] During the process of training a multi-parameter lung cancer auxiliary examination model, the in vivo imaging results of mouse lungs are morphologically characterized to construct a similarity index dataset, and this dataset is used to train a support vector machine algorithm model. This dataset is divided into a training set and a test set according to the leave-one-out method. The training set is used for the construction and training of the discrimination model, and the test set is used to test the discrimination ability of the model for lung cancer examination and optimize the detection model. As Figure 4 (e) shows, the AUC value result of classification using the support vector machine algorithm is good. Figure 4 (f) records that the classification model has achieved high classification accuracy in both original classification and cross-validation classification.

[0076] It is worth mentioning that this method uses in situ in vivo MPM images for morphological analysis of the extracellular matrix fibrous structure, which is a new idea for lung cancer auxiliary diagnosis. We use a nanoindentation test device to measure the hardness of ex vivo decellularized lung tissue and establish an association between it and the morphological similarity index of the imaging results, demonstrating the feasibility of using the morphological similarity index to reflect tissue sclerosis to identify the evolution process of lung tumors. As Figure 4 (a) shows, the hardness of ex vivo human lung cancer tissue is significantly higher than that of healthy tissue, which preliminarily proves the effectiveness of using hardness to distinguish lung cancer tissue from healthy lung tissue. Further, Figure 4 (b) The linear regression results show that the statistical indicators of the mean and standard deviation derived from the similarity index parameters can linearly reflect the hardness changes of lung tissue during the cancer development process, which provides a basis for quantitatively discriminating lung cancer tissue. Further, as Figure 4 (c) shows, the combination of the full-stack mean value, full-stack standard deviation value, and full-stack depth-direction variation value of the similarity index based on in vivo MPM images of mouse lung cancer has a high discrimination degree for different evolution process regions of lung cancer, can indirectly reflect the hardness changes of lung tissue, and realizes the extraction of information such as boundaries during the lung cancer evolution process.

[0077] Based on this idea, through steps such as optical imaging, image segmentation, morphological quantitative characterization, and constructing a similarity index classification model, the present invention finally obtains a lung cancer auxiliary examination tool based on morphological analysis of the fibrous structure of biological tissues. This method can quickly process and analyze in vivo imaging results clinically, accurately analyze the morphological status of the fibrous structure of biological tissues, realize accurate classification of healthy, boundary, and cancerous samples, and has important significance in the auxiliary examination of diseases.

[0078] The above description of the embodiments is provided to enable those of ordinary skill in the art to understand and apply the present invention. It is obvious that those who are familiar with the technology in this field can easily make various modifications to the above embodiments, and apply the general principles described herein to other embodiments without creative efforts. Therefore, the present invention is not limited to the above embodiments, and all improvements and modifications made by those skilled in the art to the present invention according to the disclosure of the present invention should be within the protection scope of the present invention.

Claims

1. A lung cancer auxiliary examination tool based on morphological analysis of biological tissue fibrous structure, characterized in that: By analyzing the morphology of the fibrous structure of biological tissues in the extracellular matrix through multiphoton microscopic imaging, a lung cancer auxiliary examination model is established. The examination tool comprises the following steps: 1) Through multiphoton microscopy, two-photon excitation fluorescence imaging and second harmonic generation imaging are performed on elastic fibers and collagen fibers in the extracellular matrix of human lung tissue to obtain multiphoton microscopy images, including elastic fiber images and collagen fiber images; 2) performing morphological multi-parameter quantitative characterization on the elastic fiber image and collagen fiber image obtained in step 1) respectively, wherein the morphological multi-parameters include local density, spatial orientation, directional variance, curvature and diameter, so as to extract the morphological characteristics of elastic fibers and collagen fibers at pixel-level accuracy. For the elastic fiber image, a series of original morphological parameter maps of elastic fibers are generated, including: elastic fiber local density parameter map, elastic fiber spatial orientation parameter map, elastic fiber directional variance parameter map, elastic fiber curvature parameter map and elastic fiber diameter parameter map. For the collagen fiber image, a series of original morphological parameter maps of collagen fibers are generated, including: collagen fiber local density parameter map, collagen fiber spatial orientation parameter map, collagen fiber directional variance parameter map and collagen fiber curvature parameter map; 3) performing image segmentation on the elastic fiber image and collagen fiber image obtained in step 1), respectively, and the image segmentation results are recorded as m1 and m2, respectively. In the segmentation results, the pixels representing the elastic fiber or collagen fiber are 1, and the pixels representing the background are 0; 4) For m1, perform cross-channel nearest neighbor fiber pixel retrieval, that is, find the nearest non-zero pixel in m2 that is away from all non-zero pixels therein, and calculate the distance between corresponding pixels to generate an elastic fiber nearest neighbor distance parameter map, and at the same time assign a value of the morphological parameter at the non-zero pixel of m1 corresponding to the nearest non-zero pixel of m2, thereby generating an elastic fiber nearest neighbor pixel morphological parameter map involving the parameters in step 2), including: an elastic fiber nearest neighbor pixel local density parameter map, an elastic fiber nearest neighbor pixel spatial orientation parameter map, an elastic fiber nearest neighbor pixel directional variance parameter map, and an elastic fiber nearest neighbor pixel curvature parameter map; 5) For m2, perform cross-channel nearest neighbor fiber pixel retrieval, that is, find the nearest non-zero pixel in m1 that is away from all non-zero pixels therein, and calculate the distance between corresponding pixels to generate a collagen fiber nearest neighbor distance parameter map, and at the same time assign a value of the morphological parameter at the non-zero pixel of m2 corresponding to the nearest non-zero pixel of m1, thereby generating a collagen fiber nearest neighbor pixel morphological parameter map involving the parameters in step 2), including: a collagen fiber nearest neighbor pixel local density parameter map, a collagen fiber nearest neighbor pixel spatial orientation parameter map, a collagen fiber nearest neighbor pixel directional variance parameter map, and a collagen fiber nearest neighbor pixel curvature parameter map; 6) performing differentiation processing on the elastic fiber nearest neighbor pixel morphological parameter map obtained in step 4) and the corresponding elastic fiber original morphological parameter map obtained in step 2) to obtain an elastic fiber nearest neighbor pixel morphological parameter difference map; 7) According to the collagen fiber nearest neighbor pixel morphological parameter map obtained in step 5), it is differentiated from the corresponding collagen fiber original morphological parameter map obtained in step 2) to obtain a collagen fiber nearest neighbor pixel morphological parameter difference map; 8) The elastic fiber nearest neighbor distance parameter map obtained in step 4) and the collagen fiber nearest neighbor distance parameter map obtained in step 5) are fused pairwise by weighting with the image segmentation result to obtain a nearest neighbor pixel distance parameter difference fusion map, and the elastic fiber nearest neighbor pixel morphological parameter difference map obtained in step 6) and the collagen fiber nearest neighbor pixel morphological parameter difference map obtained in step 7) are fused pairwise by weighting with the image segmentation result according to the corresponding morphological parameters to obtain a series of nearest neighbor pixel morphological parameter difference fusion maps, the nearest neighbor pixel distance parameter difference fusion map and the series of nearest neighbor pixel morphological parameter difference fusion maps are collectively referred to as nearest neighbor pixel parameter difference fusion maps; 9) The nearest neighbor pixel parameter difference fusion map obtained in step 8) is normalized and scored to obtain a series of normalized nearest neighbor pixel parameter difference fusion score maps, including: a normalized nearest neighbor distance parameter difference fusion score map, a normalized nearest neighbor pixel local density parameter difference fusion score map, a normalized nearest neighbor pixel spatial orientation parameter difference fusion score map, a normalized nearest neighbor pixel directional variance parameter difference fusion score map and a normalized nearest neighbor pixel curvature parameter difference fusion score map. The closer the score of the pixel in the map is to 0, the greater the difference in the corresponding morphological features of the elastic fiber and the collagen fiber at that position. The closer the score of the pixel is to 1, the higher the similarity of the corresponding morphological features of the elastic fiber and the collagen fiber at that position. Then, the corresponding pixels of all the normalized nearest neighbor pixel parameter difference fusion score maps are multiplied to finally obtain a normalized similarity index map reflecting the morphological similarity between the elastic fiber and the collagen fiber. 10) The full-stack average value, full-stack standard deviation value, and full-stack depth-direction variation value of the non-background pixel part in the normalized similarity index map are calculated respectively. The combination of the three values ​​indirectly reflects the hardness change of the lung tissue, thereby realizing the extraction of information related to the evolution of lung cancer.

2. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In the step 6), according to the elastic fiber nearest neighbor pixel morphological parameter map obtained in step 4), it is differentiated from the corresponding elastic fiber original morphological parameter map obtained in step 2), specifically: the elastic fiber nearest neighbor pixel local density parameter map is differentiated from the elastic fiber local density parameter map to obtain the elastic fiber nearest neighbor pixel local density parameter difference map, the elastic fiber nearest neighbor pixel spatial orientation parameter map is differentiated from the elastic fiber spatial orientation parameter map to obtain the elastic fiber nearest neighbor pixel spatial orientation parameter difference map, the elastic fiber nearest neighbor pixel directional variance parameter map is differentiated from the elastic fiber directional variance parameter map to obtain the elastic fiber nearest neighbor pixel spatial orientation parameter difference map The elastic fiber nearest neighbor pixel directional variance parameter difference map, the elastic fiber nearest neighbor pixel curvature parameter map and the elastic fiber curvature parameter map are differentiated to obtain the elastic fiber nearest neighbor pixel curvature parameter difference map. The elastic fiber nearest neighbor pixel local density parameter difference map, the elastic fiber nearest neighbor pixel spatial orientation parameter difference map, the elastic fiber nearest neighbor pixel directional variance parameter difference map and the elastic fiber nearest neighbor pixel curvature parameter difference map obtained by the differentiation process are collectively referred to as the elastic fiber nearest neighbor pixel morphological parameter difference map. For the morphological parameter of spatial orientation, when the difference result is between 90° and 180°, its complementary angle is taken as the differentiation result.

3. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In the step 7), according to the collagen fiber nearest neighbor pixel morphological parameter map obtained in step 5), it is differentiated from the corresponding collagen fiber original morphological parameter map obtained in step 2), specifically: the collagen fiber nearest neighbor pixel local density parameter map is differentiated from the collagen fiber local density parameter map to obtain the collagen fiber nearest neighbor pixel local density parameter difference map, the collagen fiber nearest neighbor pixel spatial orientation parameter map is differentiated from the collagen fiber spatial orientation parameter map to obtain the collagen fiber nearest neighbor pixel spatial orientation parameter difference map, the collagen fiber nearest neighbor pixel directional variance parameter map is differentiated from the collagen fiber directional variance parameter map to obtain the collagen fiber The fibril nearest neighbor pixel directional variance parameter difference map, the collagen fiber nearest neighbor pixel curvature parameter map and the collagen fiber curvature parameter map are differentiated to obtain the collagen fiber nearest neighbor pixel curvature parameter difference map. The collagen fiber nearest neighbor pixel local density parameter difference map, the collagen fiber nearest neighbor pixel spatial orientation parameter difference map, the collagen fiber nearest neighbor pixel directional variance parameter difference map and the collagen fiber nearest neighbor pixel curvature parameter difference map obtained by the differentiation process are collectively referred to as the collagen fiber nearest neighbor pixel morphological parameter difference map. For the morphological parameter of spatial orientation, when the difference result is between 90° and 180°, its complementary angle is taken as the differentiation result.

4. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In the step 8), the elastic fiber nearest neighbor pixel morphological parameter difference map obtained in step 6) and the collagen fiber nearest neighbor pixel morphological parameter difference map obtained in step 7) are fused pairwise with weighting by the image segmentation result according to the corresponding morphological parameters, specifically: the elastic fiber nearest neighbor pixel local density parameter difference map is fused with the collagen fiber nearest neighbor pixel local density parameter difference map to obtain the nearest neighbor pixel local density parameter difference fusion map, the elastic fiber nearest neighbor pixel spatial orientation parameter difference map is fused with the collagen fiber nearest neighbor pixel spatial orientation parameter difference map to obtain the nearest neighbor pixel spatial orientation parameter difference fusion map, the elastic fiber nearest neighbor pixel directional variance parameter difference map is fused with the collagen fiber nearest neighbor pixel directional variance parameter difference map to obtain the nearest neighbor pixel directional variance parameter difference fusion map, and the elastic fiber nearest neighbor pixel curvature parameter difference map is fused with the collagen fiber nearest neighbor pixel curvature parameter difference map to obtain the nearest neighbor pixel curvature parameter difference fusion map.

5. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In the step 8), when the elastic fiber nearest neighbor pixel morphological parameter difference map obtained in step 6) and the collagen fiber nearest neighbor pixel morphological parameter difference map obtained in step 7) are fused pairwise by weighting the image segmentation results according to the corresponding morphological parameters, the specific weighting method by the image segmentation results is as follows: Among them, M FP is the fusion atlas of the difference of the nearest neighbor pixel morphological parameters corresponding to the morphological parameters of elastic fibers and collagen fibers, M SP1 is the difference map of the nearest neighbor pixel morphological parameters of any morphological parameter of elastic fibers, M SP2 It is the difference map of the nearest neighbor pixel morphological parameters corresponding to the morphological parameters of collagen fibers. m1 and m2 are the image segmentation results of elastic fiber image and collagen fiber image respectively.

6. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In step 9), the nearest neighbor pixel parameter difference fusion map obtained in step 8) is normalized and scored, and the nearest neighbor distance parameter difference fusion map M is obtained. FD , using the following expression: Among them, S D is the normalized nearest neighbor distance parameter difference fusion score map, D uplimit is a reference value related to the nearest neighbor pixel distance parameter fusion map or elastic fiber diameter, which is taken as the maximum diameter value in the elastic fiber diameter parameter map, and e is the base of the natural logarithm; for the nearest neighbor pixel local density parameter difference fusion map M FL , using the following expression: Among them, S L is the normalized nearest neighbor pixel local density parameter difference fusion score map; for the nearest neighbor pixel spatial orientation parameter difference fusion map M FO , using the following expression: S O =cos(M FO ); Among them, S O It is the normalized nearest neighbor pixel spatial orientation parameter difference fusion score map; for the nearest neighbor distance direction variance parameter difference fusion map or the nearest neighbor distance curvature parameter difference fusion map M FCD , using the following expression: Among them, S CD It is the normalized nearest neighbor pixel direction variance parameter difference fusion score map or the normalized nearest neighbor pixel curvature parameter difference fusion score map. C is an empirical coefficient, which is generally taken as 4.

7. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In the step 9), the method for calculating the full-stack depth-direction variation value is as follows: firstly, the average value of all non-background pixels in each layer of the two-dimensional map of the normalized similarity index map is calculated, and then the standard deviation value of these average values ​​is calculated.

8. The lung cancer auxiliary examination tool based on biological tissue fibrous structure morphological analysis according to claim 1, characterized in that: In step 9), the information related to lung cancer evolution is extracted, and a support vector machine method is further used as an auxiliary inspection tool for lung tumor boundaries.

Images